OCR image-only PDFs

[seth-shaw-unlv]

Stemming from on an email discussion:

Hypercube currently uses pdftotext to extract text embedded in a PDF OR tesseract to perform OCR on images. However, if a user uploads a scanned document as a PDF, it won't perform OCR on the scanned document resulting in no output.

Tesseract can't process PDFs natively (ergo the pdftotext) but we can use pdfimages to extract the images into a temporary directory and loop tesseract over those to produce our extracted text OCR.

[DonRichards]

Is this still an issue?

[seth-shaw-unlv]

I think so. Tagging as a feature request until someone can confirm we have PDF OCR.

[DonRichards]

It should be working. Our installation of it has it working

[seth-shaw-unlv]

@DonRichards, and it isn't a Born-Digital specific feature? Can you test on a vanilla Islandora install to confirm?

[DonRichards]

I think this is still a problem after testing on my local.

[rosiel]

This is still an issue. When Hypercube gets a PDF, it uses pdftotext instead. This was done as part of the RDM work.

We did not realize that you can get text out of most text-containing files (and if you want into the solr index) with https://www.drupal.org/project/file_extractor.

I have a PR coming after I run the tests.

[rosiel]

I do not have a PR coming. It turns out Hypercube on its own does not accept PDFs, you need a wrapper like ocrmypdf.

[DonRichards]

Linked to this as well #1012

[adam-vessey]

Popped up in the islandora/islandora queue: Islandora/islandora#910